/ILAlgo

Baseline algorithms of imitation learning and reinforcment learning , written in Pytorch

Primary LanguagePythonMIT LicenseMIT

ILAlgo

Baseline algorithms of imitation learning and reinforcment learning , written in Pytorch. Code is tested on Mujoco.

Note: Code is under development. Issues, bug reports and PRs are welcome.

Install Dependency

pip install -r requirements

Run experiment

# train RL agent
python train_expert.py --config sac.yml --generate-data
# train non-active imitator
python train_il.py --config bc.yml
# train active imitator
python train_active_il.py --config dagger.yml

Watch Result

tensorboard --log_dir out/[your exp name]

Currently Implemented Algorithms

Reinforcement Learning

  1. TRPO
  2. PPO
  3. SAC
  4. TD3
  5. DDPG

Imitation Learning

  1. BC
  2. DAgger
  3. GAIL
  4. AIRL

Acknowledgement

During my implementation of IL and RL algorithms, a lot of classic open-source materials on the Internet served as good references. And I highly appreciate their author's effort. Below is a detailed list. By the way, thanks to my friends from LAMDA-RL for their helpful discussions.

RL Library

Open Source Code Repo

Blog

Tutorial